To date, the most powerful predictors of outcome in osteosarcoma treatment have been the ability to detect metastatic disease at diagnosis and the histopathologic response of the tumor to preoperative chemotherapy. Thus far, none of the laboratory characterizations (DNA index, cytogenetics, multiple drug resistance phenotype, etc.) have been determined to be as prognostic as the aforementioned clinical parameters. However, of those patients who do not appear to have metastases at diagnosis, approximately 30% do not survive, and these patients are difficult to identify at diagnosis. It is these patients, who do not behave as predicted by their histopathological classification that can potentially be helped by molecular analysis. An in vitro correlate which identifies these patients at diagnosis would obviously of great value. Among the possible molecular classification schemes are the analysis of expression of recessive oncogenes in these tumors. Recessive oncogenes are genes with recessive alleles that predispose to cancer. The complete absence of the normal gene product leads to tumorigenesis. Several lines of evidence point to the importance of the role of two recessive oncogenes, the retinoblastoma susceptibility gene (RB1) and the p53 gene in the etiology of osteosarcomas. First, molecular analysis has revealed that loss of expression of these two genes is a high frequency event in many histological types of osteosarcoma. Secondly, germline mutations in RB1 and p53 have been implicated in familial cancer syndromes in which osteosarcoma occurs with high frequency. The primary hypothesis in this proposal is that mutations in RB1 and p53 genes correlate with clinical outcome in osteosarcomas. A derivative hypothesis is that this correlation is more significant than the clinical prognostic factors. If so, an ultimate goal of this research is to develop a biologically based strategy for prognostication, stratification, and treatment.